Differentiating ERAN and MMN: an ERP Study

NEUROPHYSIOLOGY, BASIC AND CLINICAL NEUROREPORT

Differentiating ERAN and MMN: An ERP study

Stefan Koelsch,1,CA Thomas C. Gunter,1 Erich SchroÈger,2 Mari Tervaniemi,3 Daniela Sammler1,2 and Angela D. Friederici1

1Max Planck Institute of Cognitive Neuroscience, Stephanstr. 1a, D-04103 Leipzig; 2Institute of General Psychology, Leipzig, Germany; 3Cognitive Brain Research Unit, Helsinki, Finland

CACorresponding Author

Received 14 February 2001; accepted 27 February 2001

In the present study, the early right-anterior negativity (ERAN) both a longer latency, and a larger amplitude. The combined elicited by harmonically inappropriate chords during listening ®ndings indicate that ERAN and MMN re¯ect different mechan- to music was compared to the frequency mismatch negativity isms of pre-attentive irregularity detection, and that, although (MMN) and the abstract-feature MMN. Results revealed that both components have several features in common, the ERAN the amplitude of the ERAN, in contrast to the MMN, is does not easily ®t into the classical MMN framework. The speci®cally dependent on the degree of harmonic appropriate- present ERPs thus provide evidence for a differentiation of ness. Thus, the ERAN is correlated with the cognitive cognitive processes underlying the fast and pre-attentive processing of complex rule-based information, i.e. with the processing of auditory information. NeuroReport 12:1385±1389 application of music±syntactic rules. Moreover, results showed & 2001 Lippincott Williams & Wilkins. that the ERAN, compared to the abstract-feature MMN, had

Key words: Auditory processing; EEG; ERAN; ERP; MMN; Music

INTRODUCTION not a frequency MMN which is known to be elicited only Accurate pitch perception is a prerequisite for the proces- when a deviant tone, or chord, is preceded by a few sing of melodic, harmonic, and prosodic aspects of both standard tones or chords with identical frequency [8,9]. language and music. Recently, neural dynamics underlying Notably, both harmonically appropriate and inappropriate pitch processing within a musical context have been chords were consonant, major chords. It was only the extensively investigated by recording the mismatch nega- surrounding musical context that made some chords with tivity (MMN) [1,2], the right anterior-temporal negativity respect to principles and rules described by the theory of (RATN) [3], and the early right-anterior negativity (ERAN) harmony inappropriate (see below). [4±6], which are components of the auditory event-related Nevertheless, the ERAN is reminiscent of the MMN: (a) potential (ERP). The present study aimed at differentiating both ERAN and MMN have a similar time-course and MMN and ERAN. scalp-distribution, (b) the amplitudes of both ERAN and In previous studies investigating the perception of ma- MMN increase with the amount of violation and are jor±minor (i.e. Western) tonal music, harmonically inap- connected to behavioral discrimination performance, and propriate chords presented within a musical chord (c) both MMN and ERAN can be elicited pre-attentively sequence elicited a negativity in the ERP which was [5,8,9]. Moreover, in previous experiments the ERAN was maximal around 200 ms and right-anteriorly predominant elicited in a paradigm which has similarities to the audi- [4,5]. Due to its early latency with respect to the RATN, tory oddball paradigm, a paradigm often used to elicit an and due to its similarity to the early left anterior negativity MMN. (ELAN) [7], this ERP component was termed the early In the present study, chord sequences were presented to right-anterior negativity, or ERAN. With respect to their the participants, each sequence consisting of ®ve chords, functional signi®cance, the RATN and the ERAN are and one sequence directly following the other (middle of suggested to re¯ect the processing of musical syntax, Fig. 1). Most of the sequences consisted of in-key chords whereas the ELAN is known to be elicited by word only, but infrequently chords at the third or at the ®fth category violations and therefore taken to re¯ect syntactic position of the sequences were Neapolitan sixth chords. language processing. Neapolitan chords (in C major: f±a ¯at±d ¯at; Fig. 1) In the studies of Koelsch et al. [4±6], harmonically contain out-of-key notes (in C major: a ¯at and d ¯at) with inappropriate chords did not represent a physical deviancy respect to the harmonic context established by the preced- with respect to the preceding chords. Thus, the ERAN is ing in-key chords. Following the theory of harmony,

0959-4965 & Lippincott Williams & Wilkins Vol 12 No 7 25 May 2001 1385 NEUROREPORT S. KOELSCH ET AL. abstract feature MMN: priate/inappropriate' [13±16]. If so, the amplitude modulation of the ERAN (third versus ®fth position of a chord sequence) would not be due to any complex rule-based processing of harmonic information, but merely be depen- chord deviations: dent on the position within a stimulus train (i.e. within a sequence of acoustic events), regardless of the buildup of a musical context. In order to test whether the ERAN is an abstract feature MMN, participants of the present study were presented with acoustic events containing an abstract feature (tone etc. pairs raising (standard) or falling (deviant) in pitch, see Fig. 1), but not building up a context towards the end of each stimulus sequence. If the processes re¯ected in the ERAN are the same as those re¯ected in the MMN, the frequency MMN amplitude of the abstract feature MMN should, like the amplitude of the ERAN, increase towards the end of a stimulus sequence. Three blocks were conducted: an abstract feature MMN block, a block with chord sequences, Fig. 1. Examples of stimuli. In all blocks, stimuli were presented with the same time-course, loudness, and probability of deviant events and a frequency MMN block (with single tones, Fig. 1). In (deviant events are indicated by the arrows). all three blocks, stimuli were presented with the same time-course and the same probability of deviant events.

Neapolitans may be considered as a variation of the MATERIALS AND METHODS subdominant (a minor subdominant with a minor sixth Subjects: Twenty-eight right-handed and normal-hearing instead of a ®fth). Neapolitans at the ®fth position are subjects (aged 19±28 years, mean 23.4, 15 of them females) perceived as more inappropriate than when they are participated in the experiment. Subjects were non-musi- presented at the third position for two reasons: (a) because cians, i.e. they had never learned to play an instrument or of the musical context buildup, the tonal expectancies of professional singing, and they did not have any special listeners are more speci®c at the end of a sequence [10,11], musical education besides normal school education. and (b) since Neapolitans function in music- theory as subdominant variation, a Neapolitan at the third position Stimuli: All stimuli were played under computerized is fairly suitable (since a subdominant is appropriate), control via MIDI with a piano sound on a Roland JV-2080 whereas a Neapolitan at the ®fth position (where a tonic synthesizer. In all blocks, each stimulus sequence consisted chord is appropriate only) is inappropriate; it has been of ®ve events, presentation time (PT) of events 1±4 was shown that harmonically appropriate chord functions are 600 ms, of the ®fth event 1200 ms (in the ®rst block, the ®rst expected to a higher degree compared to inappropriate tone of each tone-pair had a PT of 100 ms, i.e. PT of the chords [10,11]. second tone was 500 ms (1100 ms at the ®fth position) Notably, harmonic expectancies of listeners follow prin- respectively). Deviant events (in the ®rst block falling tone- ciples which correspond to the harmonic relations of differ- pairs, in the second block Neapolitan chords, in the third ent chords (and keys, respectively). These relationships block tones with different frequency) occurred in some form the basis of the major±minor tonal system and are sequences randomly at either the third ( p 0.2) or ®fth described by the theory of harmony. The principles which position ( p 0.2). All chords and tones had the same govern harmonic expectancies have been described in loundness and the same decay of loudness, there was no detail as a hierarchy of harmonic stability [11]. In previous silent period between events or sequences; one sequence music experiments employing an experimental paradigm directly followed the other (Fig. 1). In each block, the same similar to that of the present study [4±6] it could be shown deviant events were employed at the third and ®fth that the degree of harmonic incongruity (i.e. the degree of position, i.e. deviant events were in each block on average harmonic expectancy violation) is re¯ected in the ampli- physically identical. Sequences containing a deviant event tude of the ERAN: The ERAN is smaller when elicited by were always preceded by a sequence exclusively consisting Neapolitans at the third position of a chord sequence of standards. Chords and tones were played with 55 dB compared to the ®fth position. That is, the elicitation and SPL. In each block, 255 sequences were presented, resulting amplitude modulation of the ERAN can be explained on in a block duration of 15 min. the basis of music theory, or, in other words, the ERAN may re¯ect cognitive processes which refer to a complex Task: In all blocks, subjects were playing a video-game rule system (which may be taken as musical syntax under the instruction to ignore all acoustic stimuli. [4,6,10±12]). This would contrast the MMN, which is In Block 1 (abstract feature MMN), standard stimuli known to be elicited either by physical deviance, or by were single-tone pairs raising in pitch, deviant pairs were rather simple abstract feature deviances which do not refer falling in pitch (top of Fig. 1). The pitch difference of two to a system of complex rules. tones of a pair was one semitone ( 6% frequency differ- However, although the ERAN cannot be a frequency ence). Three different tone pairs could occur at the ®rst MMN (see above), the ERAN could in principle be an position of a sequence, six at the second, six standards at MMN elicited by the abstract feature `harmonically appro- the third, three deviants at the third, three at the fourth,

1386 Vol 12 No 7 25 May 2001 DIFFERENTIATING ERAN AND MMN: AN ERP STUDY NEUROREPORT two standards at the ®fth, and three deviants at the ®fth (sampling rate 250 Hz). All EEG data were ®ltered off-line position (3/6/6(3)/3/2(3)). The MMN was measured from with a bandpass ®lter (0.25±25 Hz, 1001 points, FIR). the onset of the second tone of a pair. Artifacts caused by drifts or body movements were elimi- In block 2 (ERAN), all sequences consisted of ®ve chords nated by rejecting EEG data of all blocks whenever the that began with a tonic-chord. Chords at the second standard deviation within any 600 ms or 200 ms interval of position were tonic, subdominant, mediant, or submediant; all data . 25 ìV at any electrode. Eye artifacts were rejected at the third position: subdominant, Neapolitan chord, whenever the s.d. within any 200 ms interval of all data dominant, or dominant six-four chord; at the fourth posi- exceeded 30 ìV at either the vertical or the horizontal EOG. tion: dominant seventh chord; at the ®fth position: tonic or Baseline of ERPs was 50 to 0 ms relative to stimulus Neapolitan chord [17]. All chords were presented in differ- onset. ERPs were analyzed by repeated measures ANOVAs ent chordings (e.g. with the root, the third, the ®fth, and as univariate tests of hypotheses for within subjects effects. the seventh in the top voice), leading to a pool of 108 Two anterior regions of interest (ROIs) were computed: left different chord sequences. Importantly, the number of (mean of F3, FC3, F5, FC5, AF3, AF7) and right (mean of physically different chords possible at each position of a F4, FC4, F6, FC6, AF4, AF8). If not separately indicated, sequence equalled the number of different tone pairs of ANOVAs were conducted with factors condition (stan- Block 1 (3/6/6(deviant: 3 different Neapolitans)/3/2(devi- dard 3 deviant), position within the sequences (third 3 ®fth ant: 3 different Neapolitans)). Part-writing was according position), and hemisphere (left 3 right ROIs). to the classical rules of harmony [17]. In block 3 (frequency MMN), Standards were single RESULTS tones with a frequency of 440 Hz, deviant tones had a In the abstract feature MMN block, deviant tone pairs frequency of 496 Hz. presented at both the third and the ®fth position elicited an abstract feature MMN (Fig. 2, left). The latency of the Data analysis: The EEG was recorded with nose-refer- MMN was at both the third and the ®fth position around ence from 41 electrodes of the extended 10-20 system 160 ms (measured from the onset of the second tone of the

abstract feature MMN ERAN to chord devations frequency MMN

Ϫ3.0 F3 FZ F4 Ϫ3.0 F3 FZ F4 Ϫ3.0 F3 FZ F4

s s s

0.3 0.3 0.3

3.0 µV 3.0 µV 3.0 µV

F3 FZ F4 F3 FZ F4 F3 FZ F4

3rd 5th 3rd 5th 3rd 5th

standard Ϫ3.0 V 3.0 deviant

Fig. 2. ERPs elicited at frontal electrode sites by stimuli at the third (top row) and ®fth (second row) position, separately for the abstract feature MMN block, the block with chord sequences, and the frequency MMN block. Vertical line indicates the onset of the deviant stimulus (in the abstract feature MMN block: second tone of a tone pair). Bottom row: Potential maps of abstract feature MMN (standard subtracted from deviant), ERAN (harmonic appropriate chords subtracted from Neapolitan chords), and frequency MMN (standard subtracted from deviant), separately for third and ®fth position. Maps were calculated using the data from all 41 electrodes and interpolated over time windows from 125 to 185 ms (abstract feature MMN), 170±230 ms (ERAN) and 90±150 ms (frequency MMN), polarity inversions are bordered by thick lines. In contrast to the ERAN, the MMNs did not differ in amplitude between position 3 and 5, indicating that the ERAN re¯ects context-dependent musical processing. Moreover, the ERAN elicited at the ®fth position had a later latency, but larger amplitude than the abstract feature MMN.

Vol 12 No 7 25 May 2001 1387 NEUROREPORT S. KOELSCH ET AL. tone pair), the amplitudes of the MMN did not differ DISCUSSION between both positions (see also Fig. 3); no polarity inver- The present study investigated the pre-attentively activated sion was visible at mastoidal sites. An ANOVA conducted neural mechanisms of an auditory deviance detection by for a time window from 125 to 185 ms revealed an effect of comparing the ERP response to an abstract sound change, condition (F(1,26) 15.54, p , 0.0005), an interaction be- a change of harmonic appropriateness of chords, and a tween factors condition and hemisphere (F(1,26) 9.12, pitch change. Only the ERAN (elicited by the harmonically p , 0.006), and no interaction between factors condition incongruous chords in the second block) differed signi®- and position. cantly in amplitude between third and ®fth position of a In the second block (chord-sequences), Neapolitan sequence. Neither abstract feature MMN nor frequency chords at both the third and the ®fth position elicited an MMN (elicited in a non-musical context) showed an ERAN with a latency around 200 ms, the ERAN at the ®fth amplitude modulation between these positions. Since only position showed a clear polarity inversion at mastoidal the in-key chords of the second block built up a musical sites (Fig. 2, middle, the polarity inversion is indicated in context towards the end of each sequence, this ®nding the potential map). The ERAN was distinctly larger at the indicates that the amplitude of the ERAN is speci®cally ®fth than at the third position (see also Fig. 3). The correlated with the degree of harmonic incongruity in- amplitude of the ERAN elicited at the ®fth position was duced by a preceding musical context. The present results clearly larger than the amplitude of the abstract feature thus demonstrate that the ERAN re¯ects processing of MMN elicited at the ®fth position in the ®rst block. An auditory information that refers to a complex rule-based ANOVA for the data of the second block (time window system, namely rules inherent in the major±minor tonal 170±230 ms) revealed an effect of condition (F(1,26) system which are far more complex than those known to 38.53, p , 0.0001), an interaction between factors condition elicit a physical or an abstract feature MMN. and hemisphere (F(1,26) 10.24, p , 0.005), and an inter- Notably, the ERAN elicited at the ®fth position had a action between factors condition and position (F(1,26) longer latency, and a larger amplitude than the abstract 9.32, p , 0.006). ANOVAs with factor condition conducted feature MMN. The longer latency could easily be explained separately for the third and the ®fth position revealed an within the classical MMN framework if one assumes that effect of condition at both third and ®fth position (third the harmonic incongruities are more complex, and thus position: F(1,26) 4.74, p , 0.05; ®fth position: F(1,26) more dif®cult to differentiate than the deviant tone-pairs: 36.33, p , 0.0001). An ANOVA of the data from the ®fth The MMN is known to have a longer latency when stimuli position from blocks 1 and 2 with factors condition and are more dif®cult to differentiate [8,18]. Then, however, the block (1 3 2), testing the amplitude difference between the ERAN should also be smaller in amplitude than the MMN abstract feature MMN (125±185 ms) and the ERAN (170± [18]. This was not the case: on the contrary, the ERAN was 230 ms), revealed an interaction between the two factors distinctly larger than the abstract feature MMN, indicating (F(1,26) 12.43, p , 0.002). that the ERAN does not re¯ect the same cognitive pro- In the frequency MMN block, deviant tones elicited a cesses that underlie the MMN. The present ERPs thus frequency MMN at both the third and the ®fth position, provide evidence for a differentiation of cognitive pro- with a latency of around 100 ms, with clear polarity cesses underlying a pre-attentive processing of auditory inversion at mastoidal sites, and with virtually the same information in (a) sensory memory processes on the one amplitude at the third and ®fth position (Fig. 2, right, and side, and (b) relatively higher cognitive processing of Fig. 3). An ANOVA for the time window from 90 to 150 ms complex rule-based information on the other. The fact that revealed an effect of condition (F(1,26) 62.27, p , 0.0001), the ERAN speci®cally correlates with the processing of with no interaction between factors condition and position complex rules of major±minor tonal music justi®es the (no interaction was yielded between factors condition and particular term ERAN for the effects observed. hemisphere, although a slight right hemispheric prepon- It is important to note that MMN, ERAN, and ELAN all derance is visible in the potential maps). belong to a familiy of peri-sylvian (see also below) negativ- ities which re¯ect the processing of irregularities of auditory input. The present study supports the notion that 4 there are, however, considerable differences with respect to the cognitive processes and the neuronal structures mediat- 3 ing the processing of a physical irregularity like frequency on the one hand, and a language or music syntactic

V violation on the other. The fact that an MMN can also be

µ 2 elicited by abstract features might indicate that all components (physical MMNs, abstract feature MMN, ERAN and 1 ELAN) re¯ect stages on a continuum from rather simple 3rd 5th 3rd 5th 3rd 5th (physical) to fairly complex (syntactic) auditory feature 0 processing; this consideration might suggest an expansion abstract feature ERAN to chord frequency of the classical MMN framework. Notably, this considera- MMN devations MMN tion is supported by functional neuroanatomical ®ndings, Fig. 3. Amplitudes of abstract-feature MMN (left), ERAN (middle) and which indicate that the more simple features seem primar- MMN (right), separately for positions 3 (white bars) and 5 (gray bars). ily to be generated in (or in the close vicinity of) primary Latency of the abstract feature MMN was 160 ms, of the ERAN 200 ms, auditory cortical areas, with relatively small contributions and of the frequency MMN 100 ms. from the frontal areas [19±22]. In contrast, the processing

1388 Vol 12 No 7 25 May 2001 DIFFERENTIATING ERAN AND MMN: AN ERP STUDY NEUROREPORT of features which refer to a complex rule system (as framework might be expanded (at least with respect to the re¯ected in ERAN and ELAN) appears to involve more processing of major±minor tonal music) to the processing frontal, and less primary auditory structures [6,23] (a of complex, or syntactic, rules. Third, results support the recent study from Koelsch et al. [6] revealed that the ERAN hypothesis that processing of musical syntax as re¯ected in is mainly generated in the inferior part of BA 44 bilater- the ERAN is processed pre-attentively [5]. ally). Finally, the ERAN was (as the MMNs) elicited under a condition in which participants were instructed to ignore REFERENCES the chords, supporting the hypothesis that the ERAN 1. Tervaniemi M, Medvedev SV, Alho K et al. Hum Br Mapp 10,74±79 re¯ects, like the MMN [8], pre-attentive neural processes (2000). [5]. Taken together, the present results thus support the 2. Alain C, Achim A and Woods DL. Psychophysiology 36, 737±744 (1999). hypothesis of a high adaptability and ¯exibility of pre- 3. Patel AD, Gibson E, Ratner J et al. J Cogn Neurosci 10, 717±733 (1998). attentive processes in the human brain [4]. 4. Koelsch S, Gunter T, Friederici AD et al. J Cogn Neurosci 12, 520±541 (2000). 5. Koelsch S, SchroÈger E and Gunter T. submitted (2001). CONCLUSION 6. Koelsch S, Maess B and Fiederici AD. Neuroimage 11, 56 (2000). The present results demonstrate that the ERAN re¯ects 7. Hahne A and Friederici AD. J Cogn Neurosci 11, 194±205 (1999). cognitive operations connected to the processing of com- 8. NaÈaÈtaÈnen R. Attention and Brain Function. Hillsdale, NJ: Erlbaum, 1992. plex rule-based musical information, in contrast to the 9. SchroÈger E. Behav Res Methods Instr Comp 30, 131±145 (1998). MMN, which is known to re¯ect mainly sensory memory 10. Krumhansl C and Kessler E. Psych Rev 89, 334±368 (1982). operations. This ®nding is important for several reasons. 11. Bharucha J and Krumhansl C. Cognition 13, 63±102 (1983). 12. Swain J. Musical Languages. UK: Norton, 1997. First, it indicates that partly different neuronal processes 13. Saarinen J, Paavilainen P, SchroÈger E et al. Neuroreport 3, 1149±1151 underlie the generation of ERAN and MMN (although (1992). both components share several features). Since both MMN 14. Tervaniemi M, Maury S and NaÈaÈtaÈnen R. Neuroreport 5, 844±846 (1994). and ERAN can be elicited pre-attentively, the present data 15. Paavilainen P, Saarinen J, Tervaniemi M et al. Psychophysiology 9, provide evidence for a differentiation of fast and pre- 243±249 (1995). attentive neural mechanisms underlying auditory deviance 16. Paavilainen P, Jaramillo M and NaÈaÈtaÈnen R. Psychophysiology 35, 483±487 (1998). detection in the human brain. Second, although the ERAN 17. Hindemith P. Unterweisung im Tonsatz, 1. Theoretischer Teil. Mainz: does not easily ®t into the classical MMN framework, both Schott; 1940. components ®t into one concept if one considers that both 18. Tiitinen H, May P, Reinkainen K and NaÈaÈtaÈnen R. Nature 372,90±92 MMN and ERAN belong to a family of perisylvian (1994). negativites that mediate the processing of irregularities of 19. Alho K. Ear Hear 16, 38±51 (1995). auditory input. With this respect, the present results 20. Giard M, Perrin F and Pernier J. Psychophysiology 27, 627±640 (1990). 21. Alain C, Woods DL and Knight RT. Brain Res 812, 23±37 (1998). support the hypothesis of a strong adaptability and ¯ex- 22. Opitz B, Mecklinger A, von Cramon DY et al. Psychophysiology 36, ibility of fast and automatic cognitive processes in the 142±147 (1999). human brain, probably indicating that the classical MMN 23. Friederici AD, Wang Y, Herrmann C et al. Hum Br Map 11, 1±11 (2000).

Acknowledgement: The work was supported by the Leibniz Science Prize awarded to A.D. Friederici by the German Research Foundation. Full color maps, sound examples of the stimulation, and abstract with ®gures of the submitted article are available at www.stefan-koelsch.de

Vol 12 No 7 25 May 2001 1389